0-Second Start: Just Paste 1 Line into Your AI
Paste 'Run https://github.com/Atlas-Associates-Inc/atlas-longinus' into your AI agent—auto-execution runs in 20 languages to build an in-house defense AI for $0.
LONGINUS
By Atlas Associates · v0.6.0 · Apache-2.0
无需编写任何命令或参数,只需将 1 行提示词粘贴至 AI 代理即可自动启动只读安全审计。将 GCP × Firebase 攻击路径分析、5 大领域评分卡与 4 技术栈 20 规则代码诊断及机密自动脱敏融为一体,在断网状态下(外部通信 0 字节)100% 运行。

无论离线还是在线运行,LONGINUS 均对目标仓库保持 100% 只读(Read-Only: O_RDONLY | O_NOFOLLOW),使用 0o700 ACL + Fernet (AES-128-CBC + HMAC-SHA256) 加密隔离临时工作区,在审计结束时执行美国国防部标准 DoD 5220.22-M 3 次覆写擦除(0x00 → 0xFF → CSPRNG + fsync),并在每份蓝图与 PDF 报告卷末自动签发 ECDSA P-256 密码学签名的数据销毁证书。
0 Bytes
零外部通信与零遥测
4 Stacks × 20
跨栈规则 + STRIDE 诊断
12.9s / 5,302
13 个真实仓库极速扫描
$0 / Apache-2.0
零 npm 依赖 · 免费开源
01 · CORE CAPABILITIES
从 1 行 AI 自动执行到云攻击路径图谱、4 栈代码防御盾与多行机密脱敏,全部浓缩于零依赖单一引擎。
无论离线还是在线运行,LONGINUS 均对目标仓库保持 100% 只读(Read-Only: O_RDONLY | O_NOFOLLOW),使用 0o700 ACL + Fernet (AES-128-CBC + HMAC-SHA256) 加密隔离临时工作区,在审计结束时执行美国国防部标准 DoD 5220.22-M 3 次覆写擦除(0x00 → 0xFF → CSPRNG + fsync),并在每份蓝图与 PDF 报告卷末自动签发 ECDSA P-256 密码学签名的数据销毁证书。
Paste 'Run https://github.com/Atlas-Associates-Inc/atlas-longinus' into your AI agent—auto-execution runs in 20 languages to build an in-house defense AI for $0.
Quantitatively grades Firebase, IAM/WIF, Cloud Workloads, Vertex AI, and FinOps from 0–100 (Grade A–F).
Calculates external consulting fees displaced and outputs an actionable 30/60/90-day remediation roadmap.
Correlates Terraform, Rules, and IAM in milliseconds to pinpoint reachable breach paths and emit review-only diffs.
Detects DOM XSS, SQLi, SSRF, weak crypto, and key exposure across TS/Next.js, Flutter, Python/Go, and CI with STRIDE mapping.
Masks .env, PEM keys, and tokens into [REDACTED] before snippet extraction—proven across 5,302 files and 1,529 secrets with 0 leaks.
Audits multi-repo portfolios in ~12.9s and renders an in-memory local dashboard plus OASIS SARIF 2.1.0 CI reports.
Mechanically blocks mutating MCP calls and sanitizes prompt-injection payloads across 7 security profiles.
02 · ZERO-EXFILTRATION GUARANTEES
| 保护维度 | 运行保障 | 技术依据 |
|---|---|---|
| 1. 100% Offline Execution | Operates 100% with Wi-Fi disconnected (0 bytes of external traffic) | Attack-path graph, 4-stack STRIDE diagnostics, and portfolio audit execute entirely on local CPU. |
| 2. Zero Exfiltration | 0 telemetry events & 0 source-code snippets transmitted externally | Even optional AI prover mode sends zero source snippets and enforces store: false. |
| 3. Zero Persistence | 0 database or disk storage of scan results | The 127.0.0.1:4310 Web UI runs 100% in volatile memory (Cache-Control: no-store). |
| 4. Strictly Read-Only & DoD Wipe | O_RDONLY | O_NOFOLLOW Read-Only & DoD 5220.22-M 3-Pass Wipe + ECDSA P-256 Certificate | 0o700 ACL + Fernet encryption, 0x00 -> 0xFF -> CSPRNG + fsync wipe, and ECDSA P-256 signed Certificate of Data Destruction. |
| 5. Loopback Fortress | 127.0.0.1-only binding & cross-origin attack blocking | Enforces DNS Rebinding defense, Fetch Metadata validation, CSRF tokens, and CSP Nonce + Trusted Types. |
03 · DEPLOYMENT & SPRINT
Manual interviews and static slide decks. Institutional knowledge leaves when the contract ends.
Load LONGINUS into your own AI agent for autonomous attack-path analysis, 4-stack code shield, and CI gating.
Screen-share only redacted reports to complete priority patches, MCP Airlock setup, and CI gate activation.
运行 https://github.com/Atlas-Associates-Inc/atlas-longinusgit clone https://github.com/Atlas-Associates-Inc/longinus.git && cd longinus && node bin/longinus.mjs ui